
#! pip install -U textblob

Collecting textblob
  Downloading textblob-0.15.3-py2.py3-none-any.whl (636 kB)
     |████████████████████████████████| 636 kB 3.0 MB/s eta 0:00:01
Requirement already satisfied, skipping upgrade: nltk>=3.1 in /opt/anaconda3/lib/python3.7/site-packages (from textblob) (3.4.5)
Requirement already satisfied, skipping upgrade: six in /opt/anaconda3/lib/python3.7/site-packages (from nltk>=3.1->textblob) (1.12.0)
Installing collected packages: textblob
Successfully installed textblob-0.15.3
WARNING: You are using pip version 20.2.4; however, version 21.0.1 is available.
You should consider upgrading via the '/opt/anaconda3/bin/python -m pip install --upgrade pip' command.


from textblob import TextBlob


# Libraries
import os
import sqlite3 as sq
import pandas as pd
from collections import defaultdict
#!pip install wordcloud
from wordcloud import WordCloud, STOPWORDS, ImageColorGenerator
import warnings
warnings.filterwarnings("ignore")
import matplotlib.pyplot as plt
import re
import datetime
import numpy as np
from matplotlib import rc
import seaborn as sns
#!pip install plotnine
from statsmodels.tsa.arima_process import ArmaProcess
#!pip install pycausalimpact --user
from causalimpact import CausalImpact
from plotnine import *
from sklearn.feature_extraction.text import TfidfVectorizer, CountVectorizer
from sklearn.decomposition import NMF, LatentDirichletAllocation
from nltk.corpus import stopwords
from nltk.tokenize import word_tokenize, sent_tokenize
import nltk
nltk.download('stopwords')
stopwords.words('english')
nltk.download('punkt')
from nltk.stem.porter import PorterStemmer
import string
from textblob import TextBlob
from causalimpact import CausalImpact

[nltk_data] Downloading package stopwords to
[nltk_data]     /Users/sitebai/nltk_data...
[nltk_data]   Package stopwords is already up-to-date!
[nltk_data] Downloading package punkt to /Users/sitebai/nltk_data...
[nltk_data]   Package punkt is already up-to-date!


# Importing Data
os.chdir('/Users/sitebai/desktop/uci/winter/bana277/final_project') # Changing Directory according to computer
connection = sq.connect('database.sqlite')
cur = connection.cursor()


# Categories Dataset
cur.execute('SELECT * FROM categories')
categories = defaultdict(list)   
for row in cur.fetchall():
    categories['Podcast_ID'].append(row[0])
    categories['Category'].append(row[1])
    
categories = pd.DataFrame(categories, columns = list(categories.keys()))
categories


len(list(pd.DataFrame(categories['Category'].value_counts()).index))

26


# Podcasts Dataset
cur.execute('SELECT * FROM podcasts')
podcasts = defaultdict(list)   
for row in cur.fetchall():
    podcasts['Podcast_ID'].append(row[0])
    podcasts['Itunes_ID'].append(row[1])
    podcasts['Slug'].append(row[2])
    podcasts['Itunes_URL'].append(row[3])
    podcasts['Title'].append(row[4])
    
podcasts = pd.DataFrame(podcasts, columns = list(podcasts.keys()))
podcasts


def podname_finder(given_id):
    return list(podcasts[podcasts.Podcast_ID == given_id]['Title'])


# Reviews Dataset
cur.execute('SELECT * FROM reviews')
reviews = defaultdict(list)   
for row in cur.fetchall():
    reviews['Podcast_ID'].append(row[0])
    reviews['Review_Title'].append(row[1])
    reviews['Review_Description'].append(row[2])
    reviews['Review_Rating'].append(int(row[3]))
    reviews['Review_Created_At'].append(row[4])
    
reviews = pd.DataFrame(reviews, columns = list(reviews.keys()))
reviews


reviews.isnull().sum().to_frame()


categories.isnull().sum().to_frame()


podcasts.isnull().sum().to_frame()


plt.bar([str(x) for x in list(pd.DataFrame(reviews['Review_Rating'].value_counts()).index)], 
         list(reviews['Review_Rating'].value_counts()))
plt.title('Number of Reviews by Rating')
plt.xlabel('Review Rating')
plt.ylabel('Review Count')
plt.show()


# Transforming the Review Created At Variable
list1 = []
for time in reviews['Review_Created_At']:
    t = re.sub(r"-07:00", "", time)
    t = re.sub(r"T", " ", t)
    list1.append(t)

reviews['Review_DateTime'] = list1

# Keep rows that contain numbers
reviews = reviews[reviews['Review_DateTime'].str.contains(r'[0-9]+')]

# Transform to datetime object
for time in reviews['Review_DateTime']:
    t = datetime.datetime.strptime(time,"%Y-%m-%d %H:%M:%S") 
    datetime.datetime.timestamp(t)


# Extract year, month, hour, and weekday from DateTime
reviews["Review_DateTime"] = pd.to_datetime(reviews["Review_DateTime"])
reviews["Year"] = reviews["Review_DateTime"].dt.year
reviews["Month"] = reviews["Review_DateTime"].dt.month
reviews["Weekday"] = reviews["Review_DateTime"].dt.weekday
reviews["Hour"] = reviews["Review_DateTime"].dt.hour


# Reviews Over Time from 2006 to 2020
(ggplot(reviews, aes("Review_DateTime")) 
 + geom_area(stat = "bin", fill = "cornflowerblue") 
 + theme(axis_text_x = element_text(angle = 45)) 
 + labs(x = "Date", y = "Count", title = "Reviews Over Time"))

<ggplot: (8781001457701)>


# Extract rows between 2015-2021 and group them by review ratings
year = reviews[reviews['Year'] >= 2011]
#year.to_csv(r'/Users/KathyLee/UCI MSBA/Course Material/04 Winter Quarter/BANA277 Customer & Social Analytics/Project/ Reviews2011.csv', index = False)

# Group by Rating and Year
ratingbyyear = year.groupby(["Review_Rating", "Year"])["Review_Rating"].count()


# Percentage of Reviews Ratings by Year (2011-2020)
plt.figure(figsize = (9,6))

# Values of each group
bars1 = [ratingbyyear[1][2011], ratingbyyear[1][2012], ratingbyyear[1][2013], ratingbyyear[1][2014], ratingbyyear[1][2015], ratingbyyear[1][2016], ratingbyyear[1][2017], ratingbyyear[1][2018], ratingbyyear[1][2019], ratingbyyear[1][2020]]
bars2 = [ratingbyyear[2][2011], ratingbyyear[2][2012], ratingbyyear[2][2013], ratingbyyear[2][2014], ratingbyyear[2][2015], ratingbyyear[2][2016], ratingbyyear[2][2017], ratingbyyear[2][2018], ratingbyyear[2][2019], ratingbyyear[2][2020]]
bars3 = [ratingbyyear[3][2011], ratingbyyear[3][2012], ratingbyyear[3][2013], ratingbyyear[3][2014], ratingbyyear[3][2015], ratingbyyear[3][2016], ratingbyyear[3][2017], ratingbyyear[3][2018], ratingbyyear[3][2019], ratingbyyear[3][2020]]
bars4 = [ratingbyyear[4][2011], ratingbyyear[4][2012], ratingbyyear[4][2013], ratingbyyear[4][2014], ratingbyyear[4][2015], ratingbyyear[4][2016], ratingbyyear[4][2017], ratingbyyear[4][2018], ratingbyyear[4][2019], ratingbyyear[4][2020]]
bars5 = [ratingbyyear[5][2011], ratingbyyear[5][2012], ratingbyyear[5][2013], ratingbyyear[5][2014], ratingbyyear[5][2015], ratingbyyear[5][2016], ratingbyyear[5][2017], ratingbyyear[5][2018], ratingbyyear[5][2019], ratingbyyear[5][2020]]

totals = [i+j+k+z+g for i,j,k,z,g in zip(bars1, bars2, bars3, bars4, bars5)]
One = [i / j * 100 for i,j in zip(bars1, totals)]
Two = [i / j * 100 for i,j in zip(bars2, totals)]
Three = [i / j * 100 for i,j in zip(bars3, totals)]
Four = [i / j * 100 for i,j in zip(bars4, totals)]
Five = [i / j * 100 for i,j in zip(bars5, totals)]


# The position of the bars on the x-axis
r = [0,1,2,3,4,5,6,7,8,9]
 
# Names of group and bar width
names = ['2011', '2012', '2013', '2014', '2015', '2016', '2017', '2018', '2019', '2020']
barWidth = 1

# Create bars
plt.bar(r, One, color = 'slategrey', edgecolor = 'white', width = barWidth)
plt.bar(r, Two, bottom = One, color='lightsteelblue', edgecolor = 'white', width = barWidth)
plt.bar(r, Three, bottom = [i+j for i,j in zip(One,Two)], color = 'cornflowerblue', edgecolor = 'white', width = barWidth)
plt.bar(r, Four, bottom = [i+j+z for i,j,z in zip(One,Two,Three)], color = 'royalblue', edgecolor = 'white', width = barWidth)
plt.bar(r, Five, bottom = [i+j+z+g for i,j,z,g in zip(One, Two, Three, Four)], color = 'midnightblue', edgecolor = 'white', width = barWidth)

# Aes
plt.xticks(r, names)
plt.xlabel("Year", weight = "bold", size = 14)
plt.ylabel("Percentage", weight = "bold", size = 14)
plt.legend(['1','2','3','4','5'], fontsize = 12)
plt.title("Percentage of Review Ratings by Year", weight = "bold", size = 16)
plt.xticks(fontsize = 12)
plt.yticks(fontsize = 12)

# Show graphic
plt.show()


rvcnth = pd.DataFrame(reviews.groupby("Hour", as_index = False)["Hour"].count())
rvcnth = rvcnth.rename(columns = {"Hour": "Count"})
rvcnth["Hour"] = ['12 pm', '1 am', '2 am', '3 am', '4  am', '5  am', '6 am', '7 am', '8 am', '9 am', '10 am', '11 am', '12 am', '1 pm', '2 pm', '3 pm', '4 pm', '5 pm', '6 pm', '7 pm', '8 pm', '9 pm', '10 pm', '11 pm']

# Barplot
plt.figure(figsize = (22,13))

barplot = plt.bar(rvcnth["Hour"], rvcnth["Count"], color = "midnightblue", width = 0.7)
barplot[10].set_color('orange')
barplot[1].set_color('orange')

plt.title('Review Count by Hours', fontsize = 30, weight = "bold")
plt.xlabel('Hours', fontsize = 18, weight = "bold")
plt.ylabel('Count', fontsize = 18, weight = "bold")
plt.xticks(fontsize = 12)
plt.yticks(fontsize = 12)
plt.show()


# Average Reviews Rating by Hours
avgrt = reviews[["Hour", "Review_Rating"]]
avgrt = avgrt.groupby("Hour", as_index = False).mean()
avgrt["Hour"] = ['12 pm', '1 am', '2 am', '3 am', '4  am', '5  am', '6 am', '7 am', '8 am', '9 am', '10 am', '11 am', '12 am', '1 pm', '2 pm', '3 pm', '4 pm', '5 pm', '6 pm', '7 pm', '8 pm', '9 pm', '10 pm', '11 pm']

with plt.style.context('ggplot'):
    plt.figure(figsize = (20,10))
    plt.plot(avgrt["Hour"], avgrt["Review_Rating"], linestyle='solid', linewidth = 5, color = "cornflowerblue")
    plt.title('Average Review Rating by Hours', fontsize = 30, weight = "bold")
    plt.xlabel('Hours', fontsize = 18, weight = "bold")
    plt.ylabel('Review_Rating', fontsize = 18, weight = "bold")
    plt.xticks(fontsize = 12)
    plt.yticks(fontsize = 12)
plt.show()


reviewsbyweekday = pd.DataFrame(reviews.groupby("Weekday", as_index = False)["Weekday"].count())
reviewsbyweekday = reviewsbyweekday.rename(columns = {"Weekday": "Count"})
reviewsbyweekday["Weekday"] = ["Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat"]

# Barplot
plt.figure(figsize = (9, 6))
barplot = plt.bar(reviewsbyweekday["Weekday"], reviewsbyweekday["Count"], color = "midnightblue", width = 0.7)
barplot[2].set_color('orange')
barplot[5].set_color('orange')

plt.title('Review Count by Weekday', fontsize = 16, weight = "bold")
plt.xlabel('Weekday', fontsize = 14, weight = "bold")
plt.ylabel('Count', fontsize = 14, weight = "bold")
plt.xticks(fontsize = 12)
plt.yticks(fontsize = 12)
plt.show()


# Average Review Rating by Weekdays
week = reviews[["Weekday", "Review_Rating"]]
avgrw = week.groupby("Weekday", as_index = False).mean()
avgrw['Weekday'] = ["Sun", "Mon", "Tue", "Wed", "Thu", "Fri", "Sat"]

with plt.style.context('ggplot'):
    plt.figure(figsize = (10,8))
    plt.plot(avgrw["Weekday"], avgrw["Review_Rating"], linestyle='solid', linewidth = 5, color = "cornflowerblue")
    plt.title('Average Review Rating by Weekday', fontsize = 20, weight = "bold")
    plt.xlabel('Weekday', fontsize = 16, weight = "bold")
    plt.ylabel('Average Review_Rating', fontsize = 16, weight = "bold")
    plt.xticks(fontsize = 12)
    plt.yticks(fontsize = 12)    
plt.show()


# Review Rating over the Years
# The trend of number of rating over the study period from year 2006 to year 2020 since the year 2021 is not over yet.
no_2105 = reviews[(reviews.Year != 2021) &  (reviews.Year != 2005)]
(ggplot(no_2105.groupby('Year')['Review_Rating'].count().reset_index()) +
    aes('Year', 'Review_Rating') +
    geom_line())

<ggplot: (8780968282301)>


one_five = no_2105.groupby(['Year','Review_Rating'])['Podcast_ID'].count()
one_five_props = one_five.groupby(level = 0).apply(lambda x :  100 * x / float(x.sum()))
one_five_props = one_five_props.reset_index()
ones = one_five_props[one_five_props['Review_Rating'] == 1]
ones = ones.rename(columns = {'Podcast_ID':'proportion'})
fives = one_five_props[one_five_props['Review_Rating'] == 5]
fives = fives.rename(columns = {'Podcast_ID':'proportion'})

# Prepare the dataframe for creating line plot with dual Y axis. 
ones['proportion_5'] = fives.proportion.values


# As shown in the plot, in 2020, the proportion of rating 1 rised up to the highest point, and on the contrast, 
# the proportion of rating 5 dropped to the lowest point among past five years. 
fig,ax1 = plt.subplots()
ax2 = ax1.twinx()
ax1.plot(ones['Year'],ones['proportion'], 'g-' )
ax2.plot(ones['Year'], ones['proportion_5'],'b-')
ax1.set_xlabel('Year (2006 to 2020)')
ax1.set_ylabel('Proportion of Rating 1', color='g')
ax2.set_ylabel('Proportion of Rating 5', color='b')
plt.show()


cag_rev =  pd.merge(categories, reviews, on='Podcast_ID')
cag_rev['Review_Created_At'] = pd.to_datetime(cag_rev['Review_Created_At'],infer_datetime_format=True )
cag_rev['year'] = cag_rev['Review_Created_At'].dt.year
cag_rev['month'] = cag_rev['Review_Created_At'].dt.month
cag_rev['hour'] = cag_rev['Review_Created_At'].dt.hour


# Seeing the most popular categories being reviewed
cag_rev['Category'].value_counts()

society-culture                      256223
comedy                               233363
business                             126115
tv-film                              113585
education                            109310
arts                                  89659
religion-spirituality                 85430
kids-family                           57186
christianity                          52884
society-culture-personal-journals     48076
music                                 40353
business-careers                      37052
technology                            33225
business-investing                    25147
spirituality                          22334
arts-performing-arts                  20330
society-culture-philosophy            19683
society-culture-places-travel         17819
arts-visual-arts                      11067
arts-food                              9871
arts-design                            6675
arts-fashion-beauty                    6381
judaism                                1467
buddhism                                959
islam                                   800
hinduism                                169
Name: Category, dtype: int64


top_seven = ['society-culture',"comedy","business","tv-film","education",'arts',"religion-spirituality"]
top_seven_table = cag_rev[cag_rev['Category'].isin(top_seven)]


pandemic_year = top_seven_table[(top_seven_table.year == 2019) | (top_seven_table.year == 2020)]
pandemic_group = pandemic_year.groupby(['year','month','Review_Rating'])['Podcast_ID'].count()
with_categs = pandemic_year.groupby(['year','month','Category','Review_Rating'])['Podcast_ID'].count()
with_catehs_props = with_categs.groupby(level=[0,1,2]).apply(lambda x : 100 * x / float(x.sum()))
with_catehs_props = with_catehs_props.reset_index()
with_categ_new = with_catehs_props.rename(columns = {'Podcast_ID':'proportion'})
catehs_ones = with_categ_new[with_categ_new['Review_Rating'] == 1]
society_culture_p = catehs_ones[catehs_ones.Category == 'society-culture'].proportion.values
arts_p = catehs_ones[catehs_ones.Category == 'arts'].proportion.values
comedy_p = catehs_ones[catehs_ones.Category == 'comedy'].proportion.values
tv_film_p = catehs_ones[catehs_ones.Category == 'tv-film'].proportion.values
business_p = catehs_ones[catehs_ones.Category == 'business'].proportion.values
eductaion_p = catehs_ones[catehs_ones.Category == 'education'].proportion.values
religion_spirituality_p= catehs_ones[catehs_ones.Category == 'religion-spirituality'].proportion.values


plot_df = pd.DataFrame({'x':range(1,25),
                      'society':society_culture_p,
                      'arts':arts_p,
                       'comedy':comedy_p,
                       'tv_film': tv_film_p,
                        'business':business_p,
                        'education':eductaion_p,
                        'religion_spirituality':religion_spirituality_p
                       })


# Proprotion of Rating 1 over Year 2019 & Year 2020
plt.figure(figsize=(20,10))
plt.plot(plot_df['x'], plot_df['society'], marker='', color='grey', linewidth=2, alpha=0.4)
plt.plot(plot_df['x'], plot_df['arts'], marker='', color='blue', linewidth=2, alpha=0.4)
plt.plot(plot_df['x'], plot_df['comedy'], marker='', color='orange', linewidth=2, alpha=0.4)
plt.plot(plot_df['x'], plot_df['tv_film'], marker='', color='green', linewidth=2, alpha=0.4)
plt.plot(plot_df['x'], plot_df['business'], marker='', color='pink', linewidth=2, alpha=0.4)
plt.plot(plot_df['x'], plot_df['education'], marker='', color='red', linewidth=2, alpha=0.4)
plt.plot(plot_df['x'], plot_df['religion_spirituality'], marker='', color='purple', linewidth=2, alpha=0.4)
plt.plot([18, 18], [0, 18], 'k-', linestyle='dashed', lw=1)
plt.plot([15, 15], [0, 18], 'k-', linestyle='dashed', lw=1)
plt.text(18.3, 10, 'June 2020', fontsize = 16, color = 'red')
plt.text(12.7, 12.5, 'March 2020', fontsize = 16, color = 'red')

plt.title('Proprotion of Rating 1 over Year 2019 & Year 2020', fontsize = 18)
plt.xlabel('2019-2020', fontsize = 18)
plt.ylabel('Proportions', fontsize = 18)


num = 0
for i in plot_df.values[23][1:]:
    num += 1
    name = list(plot_df)[num]
    plt.text(24, i, name, horizontalalignment='left',fontsize = 12, color='black')


catehs_fives = with_categ_new[with_categ_new['Review_Rating'] == 5]
society_culture_f = catehs_fives[catehs_fives.Category == 'society-culture'].proportion.values
arts_f = catehs_fives[catehs_fives.Category == 'arts'].proportion.values
comedy_f = catehs_fives[catehs_fives.Category == 'comedy'].proportion.values
tv_film_f = catehs_fives[catehs_fives.Category == 'tv-film'].proportion.values
business_f = catehs_fives[catehs_fives.Category == 'business'].proportion.values
eductaion_f = catehs_fives[catehs_fives.Category == 'education'].proportion.values
religion_spirituality_f= catehs_fives[catehs_fives.Category == 'religion-spirituality'].proportion.values


plot_df_five = pd.DataFrame({'x':range(1,25),
                      'society':society_culture_f,
                      'arts':arts_f,
                       'comedy':comedy_f,
                       'tv_film': tv_film_f,
                        'business':business_f,
                        'education':eductaion_f,
                        'religion_spirituality':religion_spirituality_f
                       })


# Proprotion of Rating 5 over Year 2019 & Year 2020
plt.figure(figsize=(20,10))
plt.plot(plot_df['x'], plot_df_five['society'], marker='', color='grey', linewidth=2, alpha=0.4)
plt.plot(plot_df['x'], plot_df_five['arts'], marker='', color='blue', linewidth=2, alpha=0.4)
plt.plot(plot_df['x'], plot_df_five['comedy'], marker='', color='orange', linewidth=2, alpha=0.4)
plt.plot(plot_df['x'], plot_df_five['tv_film'], marker='', color='green', linewidth=2, alpha=0.4)
plt.plot(plot_df['x'], plot_df_five['business'], marker='', color='pink', linewidth=2, alpha=0.4)
plt.plot(plot_df['x'], plot_df_five['education'], marker='', color='red', linewidth=2, alpha=0.4)
plt.plot(plot_df['x'], plot_df_five['religion_spirituality'], marker='', color='purple', linewidth=1, alpha=0.4)
plt.plot([18, 18], [70, 95], 'k-', linestyle='dashed', lw=1)

plt.text(18.2, 85, 'June 2020', fontsize = 18, color = 'red')
#plt.text(9, 85, 'March 2020', fontsize = 12, color = 'red')

plt.title('Proprotion of Rating 5 over Year 2019 & Year 2020')
plt.xlabel('2019-2020')
plt.ylabel('Proportions')

num = 0
for i in plot_df_five.values[23][1:]:
    num += 1
    name = list(plot_df_five)[num]
    plt.text(24, i, name, horizontalalignment='left',fontsize = 12, color='black')


def podcat(given_id):
    '''
    Generalizing categories for a given podcast ID
    '''
    cat = list(categories[categories.Podcast_ID == given_id]['Category'])
    cat_dict = defaultdict(int)
    for c in cat:
        if 'arts' in c:
            cat_dict['arts'] += 1
        elif 'business' in c:
            cat_dict['business'] += 1
        elif 'society-culture' in c:
            cat_dict['society_culture'] += 1
        elif c in ['buddhism','christianity','hinduism','islam','judaism','religion-spirituality','spirituality']:
            cat_dict['religion-spirituality'] += 1
        else:
            cat_dict[c] += 1
    for k in cat_dict.keys():
        if cat_dict[k] == max(dict(cat_dict).values()):
            return k


# Adding Category to Podcasts Table
podcasts['Category'] = [podcat(i) for i in podcasts['Podcast_ID']]


# Adding Category to Reviews Table
reviews = reviews.merge(podcasts, on = 'Podcast_ID', how = 'right')[['Podcast_ID','Review_Title','Review_Description',
                                                                    'Review_Rating','Review_Created_At','Category']]


# Bar Graph for Generalized Categories
plt.barh(list(pd.DataFrame(podcasts['Category'].value_counts()).index), 
         list(podcasts['Category'].value_counts()))
plt.title('Number of Podcasts by Generalized Category')
plt.xlabel('Number of Podcasts')
plt.ylabel('Category')
plt.show()


plt.barh(list(pd.DataFrame(reviews['Category'].value_counts()).index), 
         list(reviews['Category'].value_counts()))
plt.title('Number of Reviews by General Category')
plt.xlabel('Number of Reviews')
plt.ylabel('Category')
plt.show()


# Average Review Rating by General Category
pd.DataFrame(reviews.groupby('Category').mean()).sort_values('Review_Rating')


def rating_clouds(given_rating):
    stopwords = set(STOPWORDS)
    stopwords.update(['podcast', 'listen', 'podcasts', 'listening','episode', 'listened','s',
                      'show', 'episodes', 'used', 'good', 'love', 'listener', 'will', 'really'
                     , 'like', 'people', 'don', 't', 'm', '&','ve', 're', 'one'])

    ratings_table = reviews[reviews.Review_Rating == given_rating]
    description_text = ' '.join(review for review in ratings_table.Review_Description)
    description_text = description_text.lower()
    description_text = description_text.strip()

    title_text = ' '.join(review for review in ratings_table.Review_Title)
    title_text = title_text.strip()
    
    print('Rating: {}'.format(given_rating))
    print('Word Cloud for Review Descriptions')
    plt.figure(figsize=(20, 20))
    plt.imshow(WordCloud(stopwords = stopwords).generate(description_text), interpolation = 'bilinear')
    plt.axis('off')
    plt.show()
    
    print('Word Cloud for Review Titles')
    plt.figure(figsize=(20, 20))
    plt.imshow(WordCloud(stopwords = stopwords).generate(title_text), interpolation = 'bilinear')
    plt.axis('off')
    plt.show()


for rating in sorted(set(reviews['Review_Rating'])):
    rating_clouds(rating)

Rating: 1
Word Cloud for Review Descriptions

Word Cloud for Review Titles

Rating: 2
Word Cloud for Review Descriptions

Word Cloud for Review Titles

Rating: 3
Word Cloud for Review Descriptions

Word Cloud for Review Titles

Rating: 4
Word Cloud for Review Descriptions

Word Cloud for Review Titles

Rating: 5
Word Cloud for Review Descriptions

Word Cloud for Review Titles


# Function for tokenization including stemming 
from nltk.stem.porter import PorterStemmer

# Initialize stemmer using porterstemmer
stemmer = PorterStemmer()

stopwords_list = []

stopwords = set(stopwords.words('english'))
stopwords1 = stopwords.union(set(stopwords_list)) # stopwords into list

def tokenize_stem(text):
    try: 
        regex = re.compile('[' +re.escape(string.punctuation) + '0-9\\r\\t\\n]') 
        text = regex.sub(" ", text) 
        
        tokens1 = [word_tokenize(s) for s in sent_tokenize(text)] 
        tokens = []
        for token_by_sent in tokens1:
            tokens += token_by_sent
        tokens = list(filter(lambda t: t.lower() not in stopwords1, tokens))
        new_tokens = [w for w in tokens if re.search('[a-zA-Z]', w)]
        new_tokens = [w.lower() for w in new_tokens if len(w)>=3]
        
        # Stemming 
        stems = [stemmer.stem(item) for item in new_tokens]
        return stems
    
    except TypeError as e: print(text,e)


# Function to print top words in topic modeling
def print_top_words(model, feature_names, n_top_words):
    for index, topic in enumerate(model.components_): #model.components_ gives an array (n_components, n_features)
        message = "Topic #{}:".format(index)
        message += " ".join([feature_names[i] for i in topic.argsort()[:-n_top_words - 1 :-1]])
        print(message)
        print("_"*70)


# Considering Data only after 2020 for analysis
cur.execute('SELECT * FROM reviews')
reviews = defaultdict(list)   
for row in cur.fetchall():
    reviews['Podcast_ID'].append(row[0])
    reviews['Review_Title'].append(row[1])
    reviews['Review_Description'].append(row[2])
    reviews['Review_Rating'].append(int(row[3]))
    reviews['Review_Created_At'].append(row[4])    
reviews = pd.DataFrame(reviews, columns = list(reviews.keys()))
reviews['Review_Created_At'] = pd.to_datetime(reviews['Review_Created_At'])
reviews = reviews[pd.DatetimeIndex(reviews['Review_Created_At']).year == 2020].reset_index(drop = True)
reviews['month'] = pd.DatetimeIndex(reviews['Review_Created_At']).month


# Tokenizing Description 
reviews['desc_stem'] = reviews['Review_Description'].map(tokenize_stem)
reviews.reset_index(drop=True, inplace=True)


# Creating Tf-Idf Scores for Review Descriptions
tfidf_vec = TfidfVectorizer(min_df=0.005,
                             max_features=None,
                             tokenizer=tokenize_stem,
                             ngram_range=(2, 3))
tfidf_e = reviews['Review_Description'].values
tfidf_vz = tfidf_vec.fit_transform(list(tfidf_e))


tfidf_table = dict(zip(tfidf_vec.get_feature_names(), tfidf_vec.idf_))

tfidf_table = pd.DataFrame(columns=['tfidf']).from_dict(dict(tfidf_table), orient='index')
tfidf_table.columns = ['tfidf']


# Lower the TF IDF score, the more common the word 
tfidf_table.sort_values(by = ['tfidf'], ascending = True).head(20)


# Higher the TF IDF score, the more rare the terms
tfidf_table.sort_values(by = ['tfidf'], ascending = False).head(20)


# Clustering Each Review
# NMF with TfidfVectorizer 
n_top_words = 15

nmf1 = NMF(init="nndsvd",
            n_components=20,
            max_iter=200, random_state=0)
A1 = tfidf_vz
W1 = nmf1.fit_transform(A1) 
H1 = nmf1.components_ 

tfidf_feature_names = tfidf_vec.get_feature_names()
print_top_words(nmf1, tfidf_feature_names, n_top_words)

Topic #0:love podcast podcast much absolut love podcast absolut love podcast love love love podcast great podcast alway year old much love podcast thank also love podcast keep new episod podcast make
______________________________________________________________________
Topic #1:listen podcast start listen podcast must listen enjoy listen time listen podcast everi podcast help stop listen podcast like learn much podcast get podcast one everi time like listen bing listen
______________________________________________________________________
Topic #2:look forward everi week new episod forward everi look forward everi podcast everi listen everi best friend learn much podcast alway everi singl make laugh everi day absolut love thank guy
______________________________________________________________________
Topic #3:feel like make feel like listen best friend like podcast podcast make everi time time listen make laugh everi week learn someth even though learn much seem like long time
______________________________________________________________________
Topic #4:love show absolut love new episod best friend listen show love love would love year old much love everi day great show sound like listen everi listen sinc also love
______________________________________________________________________
Topic #5:great podcast podcast love podcast great podcast help podcast keep podcast realli podcast alway podcast host podcast thank podcast like fun listen learn much podcast get sound like podcast one
______________________________________________________________________
Topic #6:love listen listen guy podcast love absolut love listen show everi week listen everi would love listen love also love love hear podcast alway much love make laugh year old
______________________________________________________________________
Topic #7:favorit podcast one favorit one favorit podcast podcast love podcast ever new episod podcast one podcast thank podcast alway make laugh everi week podcast host crime junki love hear learn much
______________________________________________________________________
Topic #8:highli recommend recommend podcast must listen learn much amaz podcast podcast great fun listen absolut love podcast help well done learn someth well research podcast host podcast amaz time listen
______________________________________________________________________
Topic #9:true crime crime podcast true crime podcast crime junki tell stori well research podcast one stori tell podcast love stop listen get enough one best new episod like podcast podcast like
______________________________________________________________________
Topic #10:everi episod listen everi everi week learn someth everi singl episod listen everi day forward everi look forward everi everi time podcast everi laugh loud must listen learn much make laugh
______________________________________________________________________
Topic #11:podcast listen listen everi listen sinc listen year everi week everi day found podcast listen love crime junki absolut love everi singl one best like podcast get enough seem like
______________________________________________________________________
Topic #12:great work keep great keep great work podcast keep great show great content podcast great love hear absolut love listen show crime junki listen guy great listen also love thank guy
______________________________________________________________________
Topic #13:good work keep good keep good work podcast keep great show thank guy love hear everi week realli like listen guy year old make laugh listen show great content learn much
______________________________________________________________________
Topic #14:best podcast podcast ever one best podcast one podcast love new episod podcast great listen everi year old everi singl everi day long time podcast make tell stori everi week
______________________________________________________________________
Topic #15:start listen start listen podcast year ago crime junki listen love listen year listen show get enough listen guy absolut love stop listen first episod learn much tell stori new episod
______________________________________________________________________
Topic #16:realli enjoy enjoy podcast enjoy listen podcast realli podcast great podcast much well done podcast host great show listen show seem like sound like love hear fun listen long time
______________________________________________________________________
Topic #17:love guy love love podcast love love love love absolut love great show love hear everi week would love listen guy everi time make laugh much love year old listen everi
______________________________________________________________________
Topic #18:great job podcast great great show tell stori well done love hear great listen found podcast great content stori tell must listen new episod everi week podcast one everi time
______________________________________________________________________
Topic #19:listen episod thank much found podcast learn much time listen everi time absolut love year old podcast help bing listen long time year ago stop listen podcast thank new episod
______________________________________________________________________


def getSubjectivity(text): # Function to get Subjectivity
    return TextBlob(text).sentiment.subjectivity


def getPolarity(text): # Function to get Polarity
    return TextBlob(text).sentiment.polarity


def getAnalysis(score):
    if score < 0:
        return "Negative"
    elif score == 0:
        return "Neutral"
    else:
        return "Positive"


def sentiment_analyzer(df, col):
    df["TextBlob_Subjectivity_{}".format(col)] = df[col].apply(getSubjectivity)
    df["TextBlob_Polarity_{}".format(col)] = df[col].apply(getPolarity)
    df["TextBlob_Analysis_{}".format(col)] = df["TextBlob_Polarity_{}".format(col)].apply(getAnalysis)


# Getting the sentiment of Review Description and Title
sentiment_analyzer(reviews, 'Review_Description')
sentiment_analyzer(reviews, 'Review_Title')


# Polarity throughout 2020
plt.figure(figsize=(9,4))
sns.lineplot(x = 'month', y = 'TextBlob_Polarity_Review_Description', data = reviews).set_title('Polarity Across 2020 (Description)')
plt.show()

plt.figure(figsize=(9,4))
sns.lineplot(x = 'month', y = 'TextBlob_Polarity_Review_Title', data = reviews).set_title('Polarity Across 2020 (Title)')
plt.show()


plt.bar(['Positive', 'Neutral', 'Negative'], list(reviews['TextBlob_Analysis_Review_Description'].value_counts()), 
            color = ['orange', 'grey', 'blue'])
plt.title('Review Description Sentiments')
plt.ylabel('Number of Reviews')
plt.xlabel('Sentiment')
plt.show()


plt.bar(['Positive', 'Neutral', 'Negative'], list(reviews['TextBlob_Analysis_Review_Title'].value_counts()), 
            color = ['orange', 'grey', 'blue'])
plt.title('Review Title Sentiments')
plt.ylabel('Number of Reviews')
plt.xlabel('Sentiment')
plt.show()


# Count of Ratings for Negative Sentiment
sns.countplot(x = 'Review_Rating', data = reviews[reviews['TextBlob_Analysis_Review_Description'] == 'Negative']).set_title('Count of Ratings for Negative Sentiment')
plt.show()


# Count of Ratings for Positive Sentiment
sns.countplot(x = 'Review_Rating', data = reviews[reviews['TextBlob_Analysis_Review_Description'] == 'Positive']).set_title('Count of Ratings for Positive Sentiment')
plt.show()


# Count of Ratings for Neutral Sentiment
sns.countplot(x = 'Review_Rating', data = reviews[reviews['TextBlob_Analysis_Review_Description'] == 'Neutral']).set_title('Count of Ratings for Neutral Sentiment')
plt.show()


# Filtering Reviews based on some common words associated with COVID
covid = reviews[reviews['Review_Description'].str.contains('pandemic' , regex=False, case=False, na=False)
                | reviews['Review_Description'].str.contains('corona' , regex=False, case=False, na=False)
                | reviews['Review_Description'].str.contains('virus' , regex=False, case=False, na=False)
                | reviews['Review_Description'].str.contains('covid' , regex=False, case=False, na=False)
                | reviews['Review_Description'].str.contains('cdc' , regex=False, case=False, na=False)
                | reviews['Review_Description'].str.contains('rona' , regex=False, case=False, na=False)
                | reviews['Review_Description'].str.contains('fauci' , regex=False, case=False, na=False)
                | reviews['Review_Description'].str.contains('wfh' , regex=False, case=False, na=False)
                | reviews['Review_Description'].str.contains('distanc' , regex=False, case=False, na=False)
                | reviews['Review_Description'].str.contains('sheep' , regex=False, case=False, na=False)
                | reviews['Review_Description'].str.contains('mask' , regex=False, case=False, na=False)
                | reviews['Review_Description'].str.contains('vaccine' , regex=False, case=False, na=False)
                | reviews['Review_Description'].str.contains('pfizer' , regex=False, case=False, na=False)
                | reviews['Review_Description'].str.contains('moderna' , regex=False, case=False, na=False)]


# Filtering Reviews based on some common words associated with the BLM Protests
blm = reviews[reviews['Review_Description'].str.contains('blm' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('black lives matter' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('blacklivesmatter' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('george floyd' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('racism' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('white' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('supremacy' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('trump' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('racist' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('brutality' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('breonna taylor' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('police' , regex=False, case=False, na=False)]


# Filtering Reviews based on some common words associated with the Election and Politics
politics = reviews[reviews['Review_Description'].str.contains('election' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('president' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('trump' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('biden' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('democrat' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('republican' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('govt' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('liberal' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('conservative' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('left' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('right' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('harris' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('pence' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('politics' , regex=False, case=False, na=False)|
              reviews['Review_Description'].str.contains('media' , regex=False, case=False, na=False)]


plt.bar(['Positive', 'Neutral', 'Negative'], list(covid['TextBlob_Analysis_Review_Description'].value_counts()), 
            color = ['orange', 'grey', 'blue'])
plt.title('Sentiments for COVID Related Data (Description)')
plt.ylabel('Number of Reviews')
plt.xlabel('Sentiment')
plt.show()

plt.bar(['Positive', 'Neutral', 'Negative'], list(covid['TextBlob_Analysis_Review_Title'].value_counts()), 
            color = ['orange', 'grey', 'blue'])
plt.title('Sentiments for COVID Related Data (Title)')
plt.ylabel('Number of Reviews')
plt.xlabel('Sentiment')
plt.show()


plt.bar(['Positive', 'Neutral', 'Negative'], list(blm['TextBlob_Analysis_Review_Description'].value_counts()), 
            color = ['orange', 'grey', 'blue'])
plt.title('Sentiments for BLM Related Data (Description)')
plt.ylabel('Number of Reviews')
plt.xlabel('Sentiment')
plt.show()

plt.bar(['Positive', 'Neutral', 'Negative'], list(blm['TextBlob_Analysis_Review_Title'].value_counts()), 
            color = ['orange', 'grey', 'blue'])
plt.title('Sentiments for BLM Related Data (Title)')
plt.ylabel('Number of Reviews')
plt.xlabel('Sentiment')
plt.show()


plt.bar(['Positive', 'Neutral', 'Negative'], list(politics['TextBlob_Analysis_Review_Description'].value_counts()), 
            color = ['orange', 'grey', 'blue'])
plt.title('Sentiments for Politics Related Data (Description)')
plt.ylabel('Number of Reviews')
plt.xlabel('Sentiment')
plt.show()

plt.bar(['Positive', 'Neutral', 'Negative'], list(politics['TextBlob_Analysis_Review_Title'].value_counts()), 
            color = ['orange', 'grey', 'blue'])
plt.title('Sentiments for Politics Related Data (Title)')
plt.ylabel('Number of Reviews')
plt.xlabel('Sentiment')
plt.show()


blm_2020 = cag_rev[cag_rev['year'] == 2020]
blm_2020_month = blm_2020.groupby(['month','Review_Rating'])['Podcast_ID'].count()
blm_2020_props = blm_2020_month.groupby(level = 0).apply(lambda x : 100 * x / float(x.sum()))
blm_2020_props = blm_2020_props.reset_index()
blm_2020_props = blm_2020_props.rename(columns = {'Podcast_ID':'proportion'})
blm_one_props = blm_2020_props[blm_2020_props['Review_Rating'] == 1]
blm_five_props = blm_2020_props[blm_2020_props['Review_Rating'] == 5]


# Data Prepration
blm = pd.read_csv("multiTimeline.csv")
blm = blm.reset_index()
blm.drop(0, inplace = True)
blm = blm.rename(columns = {'index': 'date', 'Category: All categories':'score'})
blm['date'] = pd.to_datetime(blm['date'], infer_datetime_format = True)
blm['year'] = blm['date'].dt.year
blm['month'] = blm['date'].dt.month
blm['day'] = blm['date'].dt.day
blm['score'] = blm['score'].astype('int')
avg_score_blm = blm.groupby('month')['score'].mean()
avg_score_blm = avg_score_blm.reset_index()


!pip install pycausalimpact

Requirement already satisfied: pycausalimpact in /opt/anaconda3/lib/python3.7/site-packages (0.1.1)
Requirement already satisfied: matplotlib>=2.2.3 in /opt/anaconda3/lib/python3.7/site-packages (from pycausalimpact) (3.1.1)
Requirement already satisfied: jinja2>=2.10 in /opt/anaconda3/lib/python3.7/site-packages (from pycausalimpact) (2.10.3)
Requirement already satisfied: scipy in /opt/anaconda3/lib/python3.7/site-packages (from pycausalimpact) (1.3.1)
Requirement already satisfied: statsmodels>=0.11.0 in /opt/anaconda3/lib/python3.7/site-packages (from pycausalimpact) (0.12.2)
Requirement already satisfied: numpy in /opt/anaconda3/lib/python3.7/site-packages (from pycausalimpact) (1.17.2)
Requirement already satisfied: cycler>=0.10 in /opt/anaconda3/lib/python3.7/site-packages (from matplotlib>=2.2.3->pycausalimpact) (0.10.0)
Requirement already satisfied: kiwisolver>=1.0.1 in /opt/anaconda3/lib/python3.7/site-packages (from matplotlib>=2.2.3->pycausalimpact) (1.1.0)
Requirement already satisfied: pyparsing!=2.0.4,!=2.1.2,!=2.1.6,>=2.0.1 in /opt/anaconda3/lib/python3.7/site-packages (from matplotlib>=2.2.3->pycausalimpact) (2.4.2)
Requirement already satisfied: python-dateutil>=2.1 in /opt/anaconda3/lib/python3.7/site-packages (from matplotlib>=2.2.3->pycausalimpact) (2.8.0)
Requirement already satisfied: MarkupSafe>=0.23 in /opt/anaconda3/lib/python3.7/site-packages (from jinja2>=2.10->pycausalimpact) (1.1.1)
Requirement already satisfied: patsy>=0.5 in /opt/anaconda3/lib/python3.7/site-packages (from statsmodels>=0.11.0->pycausalimpact) (0.5.1)
Requirement already satisfied: pandas>=0.21 in /opt/anaconda3/lib/python3.7/site-packages (from statsmodels>=0.11.0->pycausalimpact) (1.1.4)
Requirement already satisfied: six in /opt/anaconda3/lib/python3.7/site-packages (from cycler>=0.10->matplotlib>=2.2.3->pycausalimpact) (1.12.0)
Requirement already satisfied: setuptools in /opt/anaconda3/lib/python3.7/site-packages (from kiwisolver>=1.0.1->matplotlib>=2.2.3->pycausalimpact) (41.4.0)
Requirement already satisfied: pytz>=2017.2 in /opt/anaconda3/lib/python3.7/site-packages (from pandas>=0.21->statsmodels>=0.11.0->pycausalimpact) (2019.3)
WARNING: You are using pip version 20.2.4; however, version 21.0.1 is available.
You should consider upgrading via the '/opt/anaconda3/bin/python -m pip install --upgrade pip' command.


# The below summary shows that the BLM Protests did have a significant negative effect on the proportion of rating 1
y  = list(blm_one_props.proportion)
X  = list(avg_score_blm.score)
model_data = pd.DataFrame({'y': y, 'X':X}, columns = ['y', 'X'])

pre_period = [0,4]
post_period = [5, 11]
ci = CausalImpact(model_data, pre_period, post_period)
print(ci.summary())
print(ci.summary(output='report'))
ci.plot()

Posterior Inference {Causal Impact}
                          Average            Cumulative
Actual                    7.88               55.14
Prediction (s.d.)         6.4 (0.4)          44.77 (2.77)
95% CI                    [5.59, 7.14]       [39.15, 50.0]

Absolute effect (s.d.)    1.48 (0.4)         10.36 (2.77)
95% CI                    [0.73, 2.28]       [5.14, 15.99]

Relative effect (s.d.)    23.15% (6.18%)     23.15% (6.18%)
95% CI                    [11.48%, 35.72%]   [11.48%, 35.72%]

Posterior tail-area probability p: 0.0
Posterior prob. of a causal effect: 99.9%

For more details run the command: print(impact.summary('report'))
Analysis report {CausalImpact}


During the post-intervention period, the response variable had
an average value of approx. 7.88. By contrast, in the absence of an
intervention, we would have expected an average response of 6.4.
The 95% interval of this counterfactual prediction is [5.59, 7.14].
Subtracting this prediction from the observed response yields
an estimate of the causal effect the intervention had on the
response variable. This effect is 1.48 with a 95% interval of
[0.73, 2.28]. For a discussion of the significance of this effect,
see below.


Summing up the individual data points during the post-intervention
period (which can only sometimes be meaningfully interpreted), the
response variable had an overall value of 55.14.
By contrast, had the intervention not taken place, we would have expected
a sum of 44.77. The 95% interval of this prediction is [39.15, 50.0].


The above results are given in terms of absolute numbers. In relative
terms, the response variable showed an increase of +23.15%. The 95%
interval of this percentage is [11.48%, 35.72%].


This means that the positive effect observed during the intervention
period is statistically significant and unlikely to be due to random
fluctuations. It should be noted, however, that the question of whether
this increase also bears substantive significance can only be answered
by comparing the absolute effect (1.48) to the original goal
of the underlying intervention.


The probability of obtaining this effect by chance is very small
(Bayesian one-sided tail-area probability p = 0.0).
This means the causal effect can be considered statistically
significant.


# The below summary shows that the BLM Protests did have a significant negative effect on the proportion of rating 5
y1 =  list(blm_five_props.proportion)
X1 = list(avg_score_blm.score)
model_props_five = pd.DataFrame({'y':y1, 'X':X1}, columns = ['y', 'X'])
pre_period = [0,4]
post_period = [5, 11]
ci_five =CausalImpact(model_props_five, pre_period, post_period)
print(ci_five.summary())
print(ci_five.summary(output='report'))
ci_five.plot()

# After running the similar model on the covid19, 
# it seems that this shock doesn't have signifiance effect either of proportion of rating.

Posterior Inference {Causal Impact}
                          Average            Cumulative
Actual                    84.17              589.19
Prediction (s.d.)         85.65 (0.34)       599.54 (2.39)
95% CI                    [85.01, 86.35]     [595.06, 604.42]

Absolute effect (s.d.)    -1.48 (0.34)       -10.35 (2.39)
95% CI                    [-2.18, -0.84]     [-15.23, -5.87]

Relative effect (s.d.)    -1.73% (0.4%)      -1.73% (0.4%)
95% CI                    [-2.54%, -0.98%]   [-2.54%, -0.98%]

Posterior tail-area probability p: 0.0
Posterior prob. of a causal effect: 100.0%

For more details run the command: print(impact.summary('report'))
Analysis report {CausalImpact}


During the post-intervention period, the response variable had
an average value of approx. 84.17. By contrast, in the absence of an
intervention, we would have expected an average response of 85.65.
The 95% interval of this counterfactual prediction is [85.01, 86.35].
Subtracting this prediction from the observed response yields
an estimate of the causal effect the intervention had on the
response variable. This effect is -1.48 with a 95% interval of
[-2.18, -0.84]. For a discussion of the significance of this effect,
see below.


Summing up the individual data points during the post-intervention
period (which can only sometimes be meaningfully interpreted), the
response variable had an overall value of 589.19.
By contrast, had the intervention not taken place, we would have expected
a sum of 599.54. The 95% interval of this prediction is [595.06, 604.42].


The above results are given in terms of absolute numbers. In relative
terms, the response variable showed a decrease of -1.73%. The 95%
interval of this percentage is [-2.54%, -0.98%].


This means that the negative effect observed during the intervention
period is statistically significant.
If the experimenter had expected a positive effect, it is recommended
to double-check whether anomalies in the control variables may have
caused an overly optimistic expectation of what should have happened
in the response variable in the absence of the intervention.


The probability of obtaining this effect by chance is very small
(Bayesian one-sided tail-area probability p = 0.0).
This means the causal effect can be considered statistically
significant.


cag_rev_pod = pd.merge(cag_rev, podcasts, on = 'Podcast_ID')
june_2020 = cag_rev_pod[(cag_rev_pod.year == 2020) & (cag_rev_pod.month == 6)]
all_cats = june_2020.groupby(['Category_x','Review_Rating'])['Podcast_ID'].count()
all_cats_props = all_cats.groupby(level = 0).apply(lambda x : 100 * x / float(x.sum()))
all_cats_props = all_cats_props.reset_index()
sorted_one = all_cats_props[all_cats_props['Review_Rating'] == 1].sort_values(by = 'Podcast_ID', ascending = False)
sorted_five = all_cats_props[all_cats_props['Review_Rating'] == 5].sort_values(by = 'Podcast_ID', ascending = False)
top_six_june = list(sorted_one.Category_x)[0:6]
top_six = june_2020[june_2020['Category_x'].isin(top_six_june)]


def cat_clouds(category):
    stopwords = set(STOPWORDS)
    stopwords.update(['podcast', 'listen', 'podcasts', 'listening','episode', 'listened','s',
                      'show', 'episodes', 'used', 'good', 'love', 'listener', 'will', 'really'
                     , 'like', 'people', 'don', 't', 'm', '&','ve', 're', 'one'])

    ratings_table = top_six[(top_six.Review_Rating == 1) & (top_six.Category_x == category)]
    description_text = ' '.join(review for review in ratings_table.Review_Description)
    description_text = description_text.lower()
    description_text = description_text.strip()

    title_text = ' '.join(review for review in ratings_table.Review_Title)
    title_text = title_text.strip()
    
    print('Rating: {}'.format(1))
    print('Word Cloud for ' + category)
    plt.figure(figsize=(20, 20))
    plt.imshow(WordCloud(stopwords = stopwords).generate(description_text), interpolation = 'bilinear')
    plt.axis('off')
    plt.show()
    
    print('Word Cloud for Review Titles')
    plt.figure(figsize=(20, 20))
    plt.imshow(WordCloud(stopwords = stopwords).generate(title_text), interpolation = 'bilinear')
    plt.axis('off')
    plt.show()


for cat in top_six_june:
    cat_clouds(cat)

Rating: 1
Word Cloud for arts-fashion-beauty

Word Cloud for Review Titles

Rating: 1
Word Cloud for tv-film

Word Cloud for Review Titles

Rating: 1
Word Cloud for technology

Word Cloud for Review Titles

Rating: 1
Word Cloud for comedy

Word Cloud for Review Titles

Rating: 1
Word Cloud for society-culture

Word Cloud for Review Titles


ratings_table = top_six[(top_six.Review_Rating == 1) & (top_six.Category_x == 'tv-film')]
tv_film = ratings_table.groupby('Podcast_ID')['Review_Rating'].count()
tv_film = tv_film.reset_index()
tv_film['Podcast_Name'] = [podname_finder(i)[0] for i in tv_film.Podcast_ID]
tv_film.sort_values(by = 'Review_Rating', ascending =False)


pod1 = ratings_table[ratings_table['Podcast_ID'] == 'ef47233dcfdc0bf67681db6322b57f58']
pod2 = ratings_table[ratings_table['Podcast_ID'] == 'b7b4bdc0af05e557dc772b52be8cd986']


pod1.head()[['Review_Title','Review_Description']]


pod2.head()[['Review_Title','Review_Description']]


# Review with the highest positive sentiment (Same for both Titles and Descriptions)
reviews[reviews['TextBlob_Polarity_Review_Description'] == 1].head()


# Review with the neutral sentiment (Same for both Titles and Descriptions)
reviews[reviews['TextBlob_Polarity_Review_Description'] == 0].tail(10)


# Review Descriptions with the highest negative sentiment 
reviews[reviews['TextBlob_Polarity_Review_Description'] == -1].head()


# Review Descriptions with the highest negative sentiment 
reviews[reviews['TextBlob_Polarity_Review_Title'] == -1].head()


# Sentiment over the week for Descriptions
reviews['weekday'] = reviews['Review_Created_At'].dt.dayofweek
avg_sentiment = reviews.groupby(['weekday'])['TextBlob_Polarity_Review_Description'].mean().reset_index(name='Weekday Avg')
sns.lineplot(x='weekday', y='Weekday Avg', data = avg_sentiment)

<matplotlib.axes._subplots.AxesSubplot at 0x7fc68fdc3ed0>


# Sentiment over the week for Titles
avg_sentiment = reviews.groupby(['weekday'])['TextBlob_Polarity_Review_Title'].mean().reset_index(name='Weekday Avg')
sns.lineplot(x='weekday', y='Weekday Avg', data = avg_sentiment)

<matplotlib.axes._subplots.AxesSubplot at 0x7fc6907920d0>


# Sentiment over the year for titles
avg_sentiment = reviews.groupby(['month'])['TextBlob_Polarity_Review_Description'].mean().reset_index(name='Monthly Avg')
sns.lineplot(x='month', y='Monthly Avg', data = avg_sentiment)

<matplotlib.axes._subplots.AxesSubplot at 0x7fc6a0fe6190>


# Sentiment over the year for titles
avg_sentiment = reviews.groupby(['month'])['TextBlob_Polarity_Review_Title'].mean().reset_index(name='Monthly Avg')
sns.lineplot(x='month', y='Monthly Avg', data = avg_sentiment)

<matplotlib.axes._subplots.AxesSubplot at 0x7fc68fdbfd10>


# Sentiment over the day for Descriptions
reviews['hour'] = reviews['Review_Created_At'].dt.hour
avg_sentiment = reviews.groupby(['hour'])['TextBlob_Polarity_Review_Description'].mean().reset_index(name='Hourly Avg')
sns.lineplot(x='hour', y='Hourly Avg', data = avg_sentiment)

<matplotlib.axes._subplots.AxesSubplot at 0x7fc6acbefcd0>


reviews['hour'] = reviews['Review_Created_At'].dt.hour
avg_sentiment = reviews.groupby(['hour'])['TextBlob_Polarity_Review_Title'].mean().reset_index(name='Hourly Avg')
sns.lineplot(x='hour', y='Hourly Avg', data = avg_sentiment)

<matplotlib.axes._subplots.AxesSubplot at 0x7fc69fba4450>


reviews = pd.merge(reviews, podcasts, on='Podcast_ID')


for cat in sorted(set(reviews.Category)):
    cat_df = reviews[reviews.Category == cat]
    sentiment_analyzer(cat_df, 'Review_Description')
    sentiment_analyzer(cat_df, 'Review_Title')
    print('Category: {}'.format(cat.upper()))

    plt.bar(['Positive', 'Neutral', 'Negative'], list(cat_df['TextBlob_Analysis_Review_Description'].value_counts()), 
            color = ['orange', 'grey', 'blue'])
    plt.title('Category: {} [Review Description]'.format(cat.upper()))
    plt.ylabel('Number of Reviews')
    plt.xlabel('Sentiment for Review Descriptions')
    plt.show()
    
    plt.bar(['Positive', 'Neutral', 'Negative'], list(cat_df['TextBlob_Analysis_Review_Title'].value_counts()), 
            color = ['orange', 'grey', 'blue'])
    plt.title('Category: {} [Review Title]'.format(cat.upper()))
    plt.ylabel('Number of Reviews')
    plt.xlabel('Sentiment for Review Titles')
    plt.show()

Category: ARTS

Category: BUSINESS

Category: COMEDY

Category: EDUCATION

Category: KIDS-FAMILY

Category: MUSIC

Category: RELIGION-SPIRITUALITY

Category: SOCIETY_CULTURE

Category: TECHNOLOGY

Category: TV-FILM


reviews['Sentiment_Combination'] = reviews[['TextBlob_Analysis_Review_Title', 'TextBlob_Analysis_Review_Description']].values.tolist()
plt.barh([str(x) for x in list(pd.DataFrame(reviews['Sentiment_Combination'].value_counts()).index)], 
         list(reviews['Sentiment_Combination'].value_counts()))
plt.title('Number of Reviews by Sentiment Combination')
plt.xlabel('Number of Reviews')
plt.ylabel('Sentiment Combination \n[Review Title Sentiment, Review Description Sentiment]')
plt.show()

	Podcast_ID	Itunes_ID	Slug	Itunes_URL	Title
0	a00018b54eb342567c94dacfb2a3e504	1313466221	scaling-global	https://podcasts.apple.com/us/podcast/scaling-...	Scaling Global
1	a00043d34e734b09246d17dc5d56f63c	158973461	cornerstone-baptist-church-of-orlando	https://podcasts.apple.com/us/podcast/cornerst...	Cornerstone Baptist Church of Orlando
2	a0004b1ef445af9dc84dad1e7821b1e3	139076942	mystery-dancing-in-the-dark	https://podcasts.apple.com/us/podcast/mystery-...	Mystery: Dancing in the Dark
3	a00071f9aaae9ac725c3a586701abf4d	1332508972	kts-money-matters	https://podcasts.apple.com/us/podcast/kts-mone...	KTs Money Matters
4	a000aa69852b276565c4f5eb9cdd999b	1342447811	speedway-soccer	https://podcasts.apple.com/us/podcast/speedway...	Speedway Soccer
...	...	...	...	...	...
46660	fffe3f208a56dfecfaf6d0a7f8399d63	1420703219	how-travel-writers-self-publish	https://podcasts.apple.com/us/podcast/how-trav...	How Travel Writers Self-Publish
46661	fffeb7d6d05f2b4c600fbebc828ca656	1220681898	teddy-the-empress-cooking-the-queens	https://podcasts.apple.com/us/podcast/teddy-th...	TEDDY & THE EMPRESS: Cooking the Queens
46662	ffff5db4b5db2d860c49749e5de8a36d	384521934	frankenstein-or-the-modern-prometheus	https://podcasts.apple.com/us/podcast/frankens...	Frankenstein, or the Modern Prometheus
46663	ffff66f98c1adfc8d0d6c41bb8facfd0	1435731839	whos-bringing-wine	https://podcasts.apple.com/us/podcast/whos-bri...	Who’s Bringing Wine?
46664	ffff923482740bc21a0fe184865ec2e2	1271277212	tefl-waffle	https://podcasts.apple.com/us/podcast/tefl-waf...	TEFL Waffle

	Podcast_ID	Review_Title	Review_Description	Review_Rating	Review_Created_At
0	c61aa81c9b929a66f0c1db6cbe5d8548	really interesting!	Thanks for providing these insights. Really e...	5	2018-04-24T12:05:16-07:00
1	c61aa81c9b929a66f0c1db6cbe5d8548	Must listen for anyone interested in the arts!!!	Super excited to see this podcast grow. So man...	5	2018-05-09T18:14:32-07:00
2	ad4f2bf69c72b8db75978423c25f379e	nauseatingly left	I'm a liberal myself, but its pretty obvious a...	1	2019-06-11T14:53:39-07:00
3	ad4f2bf69c72b8db75978423c25f379e	Diverse stories	I find Tedx talks very inspirational but I oft...	5	2018-05-31T13:08:09-07:00
4	ad4f2bf69c72b8db75978423c25f379e	👍👍👍👍	I love this podcast, it is so good.	5	2019-06-19T13:56:05-07:00
...	...	...	...	...	...
1162835	ddd451a18055f0108edf79f8c3c9bf15	Just want everyone to have fun 😊😊😊😊😊	I made this so all you guys that listen to thi...	5	2021-02-05T04:29:16-07:00
1162836	a08a93f2ef58f847b263c84d96e891f5	Mind blowing content	I just discovered Kate Doster and the hype is ...	5	2021-02-04T14:31:55-07:00
1162837	f9255ade54a8e12df362cde70b49dff9	Way to many ads	This podcast as well as the “you can sit with ...	1	2021-02-04T17:09:19-07:00
1162838	f9255ade54a8e12df362cde70b49dff9	Meh.	Really crappy ep 2/4. I guess they can’t all b...	1	2021-02-04T17:41:06-07:00
1162839	cf6b01b278eaa01a3e7bdb607ebc3d66	Loved it!	Very well presented story! Sad to have happene...	5	2021-02-04T09:40:13-07:00

	tfidf
make podcast	6.289763
bing listen	6.284231
podcast amaz	6.277164
podcast get	6.277164
great listen	6.270146
look forward everi	6.270146
one favorit podcast	6.267043
forward everi	6.257791
absolut love podcast	6.235779
podcast keep	6.228299
start listen podcast	6.227555
episod listen	6.220136
first time	6.220136
podcast everi	6.219397
well research	6.212772
listen love	6.206191
laugh loud	6.190283
even though	6.183135
thank guy	6.171100
stori tell	6.164786

	Review_Title	Review_Description
1309960	Bye, Rachel!	So you talk trash on the show that got you to ...
1309961	Boring	Expand your topics. It’s been miserable to lis...
1309962	Bye, Rachel!	So you talk trash on the show that got you to ...
1309966	Rachel deletes most negative reviews	Rachel is a bully and is mean \n\nShe is unapp...
1309968	Rachel is a bully and a mean person	Rachel is mean \n\nAnyone find it funny that s...

	Review_Title	Review_Description
704450	She makes racist comments on her podcast	On one of her podcasts she said is over the bl...
704451	Racism	Stassi and he costars on Vanderpump rules inte...
704452	Stassi is trash	Never forget she and Kristen called the cops t...
704453	Lame	I used to really like Stassi when her podcast ...
704454	Stassi needs to go away	Stassi is a self indulgent racist that is tryi...

Analysis of Podcast Reviews¶

Exploratory Data Analysis¶

By Year (2011-2020)¶

By Hour¶

By Weekday¶

Text Modelling¶

Sentiment Analysis¶

Changes During Historic Events in 2020¶

Combination and Categorical Sentiment Analysis¶

	Podcast_ID	Category
0	c61aa81c9b929a66f0c1db6cbe5d8548	arts
1	c61aa81c9b929a66f0c1db6cbe5d8548	arts-performing-arts
2	c61aa81c9b929a66f0c1db6cbe5d8548	music
3	ad4f2bf69c72b8db75978423c25f379e	arts
4	ad4f2bf69c72b8db75978423c25f379e	arts-design
...	...	...
70947	cf6e43ec4739235d68ed0b69b1604d65	society-culture
70948	b73527e94e7e0ce7fe546c818a471a2a	comedy
70949	cd2660463f3db26d95d341b74a9ffd9f	technology
70950	c52c7bc59903251a9b75b27d4ba64802	kids-family
70951	c52c7bc59903251a9b75b27d4ba64802	music

	Review_Rating
Category
society_culture	4.532122
technology	4.537932
tv-film	4.559356
comedy	4.641795
kids-family	4.690365
arts	4.747529
music	4.768538
education	4.778256
religion-spirituality	4.818149
business	4.837249

	tfidf
love podcast	3.699773
listen podcast	3.925193
look forward	4.324125
feel like	4.398882
podcast listen	4.732202
true crime	4.738565
love listen	4.760473
everi episod	4.805424
favorit podcast	4.812092
love show	4.835136
start listen	4.881524
great podcast	4.896952
highli recommend	4.916428
everi week	5.073896
listen everi	5.129687
best podcast	5.193001
realli enjoy	5.195120
new episod	5.210082
podcast love	5.240139
great work	5.244865

	Podcast_ID	Review_Rating	Podcast_Name
17	b7b4bdc0af05e557dc772b52be8cd986	123	Straight Up with Stassi
55	ef47233dcfdc0bf67681db6322b57f58	79	Bachelor Happy Hour with Rachel & Ali – The Of...
28	ce295a06b5c7d06a1d123cdc66fb70c7	26	The Ben and Ashley I Almost Famous Podcast
19	bd6efdaac64b9da221f25efd4b285352	15	Scrubbing In with Becca Tilley & Tanya Rad
35	d739564fca425b41c8c785d82ec64c18	11	Off The Vine with Kaitlyn Bristowe
...	...	...	...
30	d17a9b8d834a0eb4ca3700b63905cf1b	1	Pod Sematary
32	d42338cf59d2a6cc6c59e52878d14893	1	Insecuritea: The Insecure Aftershow
37	d8d287b2134bca088cbdeefb8b81b6ab	1	Story and Star Wars
38	da7ff73d5364e8b8b679d1b72f171202	1	Classic Movie Musts
66	fe7b01d58d1faf093f5deb9b27006ed8	1	Hallmark Channels’ Bubbly Sesh

	Podcast_ID	Review_Title	Review_Description	Review_Rating	Review_Created_At	month	desc_stem	TextBlob_Subjectivity_Review_Description	TextBlob_Polarity_Review_Description	TextBlob_Analysis_Review_Description	TextBlob_Subjectivity_Review_Title	TextBlob_Polarity_Review_Title	TextBlob_Analysis_Review_Title
45	b1a3eb2aa8e82ecbe9c91ed9a963c362	So good	Keep up the great work!! This podcast always h...	5	2020-01-01 07:22:39-07:00	1	[keep, great, work, podcast, alway, crack]	0.75	1.0	Positive	0.60	0.70	Positive
63	bf5bf76d5b6ffbf9a31bba4480383b7f	Addicting!	Best podcast! Got my mom hooked on it too!	5	2020-01-01 13:29:25-07:00	1	[best, podcast, got, mom, hook]	0.30	1.0	Positive	0.00	0.00	Neutral
104	c3f080cc393035a81d4ac7c7bff1c6c1	😀😍	Awesome 😎 \nFrom my sister:💍👰🧜🏼‍♀️🧚‍♀️👭🦄🦋💐🌸🌷🌈🏵...	5	2020-01-01 11:49:53-07:00	1	[awesom, sister]	1.00	1.0	Positive	0.00	0.00	Neutral
334	bad6c91efdbee814db985c7a65199604	I loooooooooove it.	It is sooooooooooooo good!!!!!!!!!!!	5	2020-01-02 05:47:41-07:00	1	[sooooooooooooo, good]	0.60	1.0	Positive	0.00	0.00	Neutral
335	bad6c91efdbee814db985c7a65199604	Cool 😎	Best podcast ever🙂😎🥰😍🎃	5	2020-01-02 08:57:14-07:00	1	[best, podcast, ever🙂😎🥰😍🎃]	0.30	1.0	Positive	0.65	0.35	Positive

	Podcast_ID	Review_Title	Review_Description	Review_Rating	Review_Created_At	month	desc_stem	TextBlob_Subjectivity_Review_Description	TextBlob_Analysis_Review_Description	TextBlob_Subjectivity_Review_Title	TextBlob_Polarity_Review_Title	TextBlob_Analysis_Review_Title
250136	ab6d7b6685d410d400b8d5d5feaa939e	Super Nuck	Rev. Don Wildman is one of my heroes	5	2020-12-03 09:28:26-07:00	12	[rev, wildman, one, hero]	0.000000	Neutral	0.666667	0.333333	Positive
250137	dd55eec1cef72c996f964f2422ab39b1	Amazing podcast	Thank you for this podcast!!	5	2020-12-31 11:07:35-07:00	12	[thank, podcast]	0.000000	Neutral	0.900000	0.600000	Positive
250153	e6bd9017b718f362b79c52e97591d5a7	Have some fun!	This is a simple and laid back podcast that I ...	5	2020-12-20 20:55:32-07:00	12	[simpl, laid, back, podcast, drive, listen]	0.178571	Neutral	0.200000	0.375000	Positive
250157	c4ffdcc64b1f3841a8d622494356e5ff	Positively stunning	The appreciation for queer culture? The ever-c...	5	2020-10-29 11:01:18-07:00	10	[appreci, queer, cultur, ever, chang, evolv, d...	0.000000	Neutral	1.000000	0.500000	Positive
250164	b746b49e29e790c2d192bc7e42f4f707	Gracias!	Me encanta el programa! Y me fascina la manera...	5	2020-11-13 09:59:00-07:00	11	[encanta, programa, fascina, manera, que, rela...	0.000000	Neutral	0.000000	0.000000	Neutral
250170	bb56f29307c109aac3253e46dacc2618	Thank you!	Thank you for passing on the wisdom that God h...	5	2020-11-14 21:57:08-07:00	11	[thank, pass, wisdom, god, given, encourag]	0.000000	Neutral	0.000000	0.000000	Neutral
250171	bb56f29307c109aac3253e46dacc2618	Solid, biblical truth.	Focused on Christ, focused on truth.	5	2020-12-21 20:37:47-07:00	12	[focus, christ, focus, truth]	0.000000	Neutral	0.100000	0.000000	Neutral
250212	c0d2577e62937ec804a3b0fd2b571ec1	25/09/2020 : totalement vide de sens	D’abord, merci Marc Fiorentino pour avoir semb...	1	2020-10-01 20:18:02-07:00	10	[abord, merci, marc, fiorentino, pour, avoir, ...	0.000000	Neutral	0.000000	0.000000	Neutral
250213	b234cb7562458a60593c62eec4983a4c	Thank You	You hit everything on point each episode! List...	5	2020-10-22 11:19:39-07:00	10	[hit, everyth, point, episod, listen, bring, p...	0.000000	Neutral	0.000000	0.000000	Neutral
250219	c819ca1a90a8fc459e336feacd8677b9	C’est parfait	Ce podcast est parfait pour ceux qui aiment le...	5	2020-09-24 01:09:40-07:00	9	[podcast, est, parfait, pour, ceux, qui, aimen...	0.000000	Neutral	0.000000	0.000000	Neutral

	Podcast_ID	Review_Title	Review_Description	Review_Rating	Review_Created_At	month	desc_stem	TextBlob_Subjectivity_Review_Description	TextBlob_Polarity_Review_Description	TextBlob_Analysis_Review_Description	TextBlob_Subjectivity_Review_Title	TextBlob_Polarity_Review_Title	TextBlob_Analysis_Review_Title
1	bf5bf76d5b6ffbf9a31bba4480383b7f	Crime Junkie	If it wasn’t for Brit. I would give you a 5sta...	3	2020-01-01 02:32:09-07:00	1	[brit, would, give, star, review, show, never,...	1.0	-1.0	Negative	0.000000	0.00	Neutral
1273	e2db9bd35adb4ffff176921fc29a79d3	Great show	These guys know their stuff and bring a lot of...	5	2020-01-02 07:24:16-07:00	1	[guy, know, stuff, bring, lot, info, tabl, wit...	1.0	-1.0	Negative	0.750000	0.80	Positive
1773	bc636a254a66dafef93ff5185d88e22a	Huh?	!!!!!!!!!!!!!!!!!!!!!!! What the heck is this ...	1	2020-01-02 20:20:58-07:00	1	[heck, horribl, horribl, horribl, horribl]	1.0	-1.0	Negative	0.000000	0.00	Neutral
1926	ae3aa1fed48c0e737cdb3ccfa7efc36e	Repetitive	Worst podcast i have listened to.	1	2020-01-05 07:32:06-07:00	1	[worst, podcast, listen]	1.0	-1.0	Negative	0.250000	-0.25	Negative
2813	d62617013be20649841199461f99d5c3	Cried laughing 10 times within first 10 episodes	I’m talking by myself on the side walk hysteri...	5	2020-01-06 14:25:46-07:00	1	[talk, side, walk, hyster, laughter]	1.0	-1.0	Negative	0.333333	0.25	Positive

	Podcast_ID	Review_Title	Review_Description	Review_Rating	Review_Created_At	month	desc_stem	TextBlob_Subjectivity_Review_Description	TextBlob_Polarity_Review_Description	TextBlob_Analysis_Review_Description	TextBlob_Subjectivity_Review_Title	TextBlob_Polarity_Review_Title	TextBlob_Analysis_Review_Title
95	bc5ddad3898e0973eb541577d1df8004	Awful	I tried giving it a shot due to all the hype b...	1	2020-01-01 13:52:57-07:00	1	[tri, give, shot, due, hype, found, pretti, un...	0.553571	0.014286	Positive	1.0	-1.0	Negative
141	d7e311881667310f3f72cde517c25060	I fee awful that I watched both of these shows	And I enjoyed them at the time with no thought...	5	2020-01-01 08:09:46-07:00	1	[enjoy, time, thought, back, stori, impact, li...	0.350000	0.250000	Positive	1.0	-1.0	Negative
147	bfe2ea080760d6108cd24516da2fc9ae	Boring	Great topics made to excite you but simply bor...	1	2020-01-01 17:07:09-07:00	1	[great, topic, made, excit, simpli, bore, two]	0.553571	0.400000	Positive	1.0	-1.0	Negative
325	e829449dfaf0758df2773fb9c8f3db1b	Horrible	If I could give Zero starts I would.	1	2020-01-02 04:59:36-07:00	1	[could, give, zero, start, would]	0.000000	0.000000	Neutral	1.0	-1.0	Negative
541	d4baff0192b5f556b0c3f2a273e98e93	I AM APPALLED! 🙄🤪	Long time listener and finally close to caught...	5	2020-01-02 17:09:17-07:00	1	[long, time, listen, final, close, caught, has...	0.691667	0.127604	Positive	1.0	-1.0	Negative